OLD | NEW |
| (Empty) |
1 // Copyright 2014 The Chromium Authors. All rights reserved. | |
2 // Use of this source code is governed by a BSD-style license that can be | |
3 // found in the LICENSE file. | |
4 | |
5 // The basic usage of the Filter interface is described in the comment at | |
6 // the beginning of filter.h. If Filter::Factory is passed a vector of | |
7 // size greater than 1, that interface is implemented by a series of filters | |
8 // connected in a chain. In such a case the first filter | |
9 // in the chain proxies calls to ReadData() so that its return values | |
10 // apply to the entire chain. | |
11 // | |
12 // In a filter chain, the data flows from first filter (held by the | |
13 // caller) down the chain. When ReadData() is called on any filter | |
14 // except for the last filter, it proxies the call down the chain, | |
15 // filling in the input buffers of subsequent filters if needed (== | |
16 // that filter's last_status() value is FILTER_NEED_MORE_DATA) and | |
17 // available (== the current filter has data it can output). The last | |
18 // Filter will then output data if possible, and return | |
19 // FILTER_NEED_MORE_DATA if not. Because the indirection pushes | |
20 // data along the filter chain at each level if it's available and the | |
21 // next filter needs it, a return value of FILTER_NEED_MORE_DATA from the | |
22 // final filter will apply to the entire chain. | |
23 | |
24 #include "net/filter/filter.h" | |
25 | |
26 #include "base/files/file_path.h" | |
27 #include "base/strings/string_util.h" | |
28 #include "base/values.h" | |
29 #include "net/base/io_buffer.h" | |
30 #include "net/base/sdch_net_log_params.h" | |
31 #include "net/filter/brotli_filter.h" | |
32 #include "net/filter/gzip_filter.h" | |
33 #include "net/filter/sdch_filter.h" | |
34 #include "net/log/net_log_event_type.h" | |
35 #include "net/log/net_log_with_source.h" | |
36 #include "net/url_request/url_request_context.h" | |
37 #include "url/gurl.h" | |
38 | |
39 namespace net { | |
40 | |
41 namespace { | |
42 | |
43 // Filter types (using canonical lower case only): | |
44 const char kBrotli[] = "br"; | |
45 const char kDeflate[] = "deflate"; | |
46 const char kGZip[] = "gzip"; | |
47 const char kXGZip[] = "x-gzip"; | |
48 const char kSdch[] = "sdch"; | |
49 // compress and x-compress are currently not supported. If we decide to support | |
50 // them, we'll need the same mime type compatibility hack we have for gzip. For | |
51 // more information, see Firefox's nsHttpChannel::ProcessNormal. | |
52 | |
53 // Mime types: | |
54 const char kTextHtml[] = "text/html"; | |
55 | |
56 // Buffer size allocated when de-compressing data. | |
57 const int kFilterBufSize = 32 * 1024; | |
58 | |
59 void LogSdchProblem(const FilterContext& filter_context, | |
60 SdchProblemCode problem) { | |
61 SdchManager::SdchErrorRecovery(problem); | |
62 filter_context.GetNetLog().AddEvent( | |
63 NetLogEventType::SDCH_DECODING_ERROR, | |
64 base::Bind(&NetLogSdchResourceProblemCallback, problem)); | |
65 } | |
66 | |
67 std::string FilterTypeAsString(Filter::FilterType type_id) { | |
68 switch (type_id) { | |
69 case Filter::FILTER_TYPE_BROTLI: | |
70 return "FILTER_TYPE_BROTLI"; | |
71 case Filter::FILTER_TYPE_DEFLATE: | |
72 return "FILTER_TYPE_DEFLATE"; | |
73 case Filter::FILTER_TYPE_GZIP: | |
74 return "FILTER_TYPE_GZIP"; | |
75 case Filter::FILTER_TYPE_GZIP_HELPING_SDCH: | |
76 return "FILTER_TYPE_GZIP_HELPING_SDCH"; | |
77 case Filter::FILTER_TYPE_SDCH: | |
78 return "FILTER_TYPE_SDCH"; | |
79 case Filter::FILTER_TYPE_SDCH_POSSIBLE : | |
80 return "FILTER_TYPE_SDCH_POSSIBLE "; | |
81 case Filter::FILTER_TYPE_UNSUPPORTED: | |
82 return "FILTER_TYPE_UNSUPPORTED"; | |
83 case Filter::FILTER_TYPE_MAX: | |
84 return "FILTER_TYPE_MAX"; | |
85 } | |
86 return ""; | |
87 } | |
88 | |
89 } // namespace | |
90 | |
91 FilterContext::~FilterContext() { | |
92 } | |
93 | |
94 Filter::~Filter() {} | |
95 | |
96 // static | |
97 std::unique_ptr<Filter> Filter::Factory( | |
98 const std::vector<FilterType>& filter_types, | |
99 const FilterContext& filter_context) { | |
100 if (filter_types.empty()) | |
101 return nullptr; | |
102 | |
103 std::unique_ptr<Filter> filter_list = nullptr; // Linked list of filters. | |
104 for (size_t i = 0; i < filter_types.size(); i++) { | |
105 filter_list = PrependNewFilter(filter_types[i], filter_context, | |
106 kFilterBufSize, std::move(filter_list)); | |
107 if (!filter_list) | |
108 return nullptr; | |
109 } | |
110 return filter_list; | |
111 } | |
112 | |
113 // static | |
114 std::unique_ptr<Filter> Filter::GZipFactory() { | |
115 return InitGZipFilter(FILTER_TYPE_GZIP, kFilterBufSize); | |
116 } | |
117 | |
118 // static | |
119 std::unique_ptr<Filter> Filter::FactoryForTests( | |
120 const std::vector<FilterType>& filter_types, | |
121 const FilterContext& filter_context, | |
122 int buffer_size) { | |
123 if (filter_types.empty()) | |
124 return nullptr; | |
125 | |
126 std::unique_ptr<Filter> filter_list; // Linked list of filters. | |
127 for (size_t i = 0; i < filter_types.size(); i++) { | |
128 filter_list = PrependNewFilter(filter_types[i], filter_context, buffer_size, | |
129 std::move(filter_list)); | |
130 if (!filter_list) | |
131 return nullptr; | |
132 } | |
133 return filter_list; | |
134 } | |
135 | |
136 Filter::FilterStatus Filter::ReadData(char* dest_buffer, int* dest_len) { | |
137 const int dest_buffer_capacity = *dest_len; | |
138 if (last_status_ == FILTER_ERROR) | |
139 return last_status_; | |
140 if (!next_filter_.get()) | |
141 return last_status_ = ReadFilteredData(dest_buffer, dest_len); | |
142 | |
143 // This filter needs more data, but it's not clear that the rest of | |
144 // the chain does; delegate the actual status return to the next filter. | |
145 if (last_status_ == FILTER_NEED_MORE_DATA && !stream_data_len()) { | |
146 last_status_ = next_filter_->ReadData(dest_buffer, dest_len); | |
147 return last_status_; | |
148 } | |
149 | |
150 do { | |
151 if (next_filter_->last_status() == FILTER_NEED_MORE_DATA) { | |
152 PushDataIntoNextFilter(); | |
153 if (FILTER_ERROR == last_status_) { | |
154 *dest_len = 0; | |
155 return FILTER_ERROR; | |
156 } | |
157 } | |
158 *dest_len = dest_buffer_capacity; // Reset the input/output parameter. | |
159 | |
160 next_filter_->ReadData(dest_buffer, dest_len); | |
161 if (FILTER_NEED_MORE_DATA == last_status_) { | |
162 last_status_ = next_filter_->last_status(); | |
163 return last_status_; | |
164 } | |
165 | |
166 // In the case where this filter has data internally, and is indicating such | |
167 // with a last_status_ of FILTER_OK, but at the same time the next filter in | |
168 // the chain indicated it FILTER_NEED_MORE_DATA, we have to be cautious | |
169 // about confusing the caller. The API confusion can appear if we return | |
170 // FILTER_OK (suggesting we have more data in aggregate), but yet we don't | |
171 // populate our output buffer. When that is the case, we need to | |
172 // alternately call our filter element, and the next_filter element until we | |
173 // get out of this state (by pumping data into the next filter until it | |
174 // outputs data, or it runs out of data and reports that it NEED_MORE_DATA.) | |
175 } while (FILTER_OK == last_status_ && | |
176 FILTER_NEED_MORE_DATA == next_filter_->last_status() && | |
177 0 == *dest_len); | |
178 | |
179 if (next_filter_->last_status() == FILTER_ERROR) { | |
180 last_status_ = FILTER_ERROR; | |
181 *dest_len = 0; | |
182 return FILTER_ERROR; | |
183 } | |
184 last_status_ = FILTER_OK; | |
185 return FILTER_OK; | |
186 } | |
187 | |
188 bool Filter::FlushStreamBuffer(int stream_data_len) { | |
189 DCHECK_LE(stream_data_len, stream_buffer_size_); | |
190 if (stream_data_len <= 0 || stream_data_len > stream_buffer_size_) | |
191 return false; | |
192 | |
193 DCHECK(stream_buffer()); | |
194 // Bail out if there is more data in the stream buffer to be filtered. | |
195 if (!stream_buffer() || stream_data_len_) | |
196 return false; | |
197 | |
198 next_stream_data_ = stream_buffer()->data(); | |
199 stream_data_len_ = stream_data_len; | |
200 last_status_ = FILTER_OK; | |
201 return true; | |
202 } | |
203 | |
204 // static | |
205 Filter::FilterType Filter::ConvertEncodingToType( | |
206 const std::string& filter_type) { | |
207 FilterType type_id; | |
208 if (base::LowerCaseEqualsASCII(filter_type, kBrotli)) { | |
209 type_id = FILTER_TYPE_BROTLI; | |
210 } else if (base::LowerCaseEqualsASCII(filter_type, kDeflate)) { | |
211 type_id = FILTER_TYPE_DEFLATE; | |
212 } else if (base::LowerCaseEqualsASCII(filter_type, kGZip) || | |
213 base::LowerCaseEqualsASCII(filter_type, kXGZip)) { | |
214 type_id = FILTER_TYPE_GZIP; | |
215 } else if (base::LowerCaseEqualsASCII(filter_type, kSdch)) { | |
216 type_id = FILTER_TYPE_SDCH; | |
217 } else { | |
218 // Note we also consider "identity" and "uncompressed" UNSUPPORTED as | |
219 // filter should be disabled in such cases. | |
220 type_id = FILTER_TYPE_UNSUPPORTED; | |
221 } | |
222 return type_id; | |
223 } | |
224 | |
225 // static | |
226 void Filter::FixupEncodingTypes( | |
227 const FilterContext& filter_context, | |
228 std::vector<FilterType>* encoding_types) { | |
229 std::string mime_type; | |
230 bool success = filter_context.GetMimeType(&mime_type); | |
231 DCHECK(success || mime_type.empty()); | |
232 | |
233 // If the request was for SDCH content, then we might need additional fixups. | |
234 if (!filter_context.SdchDictionariesAdvertised()) { | |
235 // It was not an SDCH request, so we'll just record stats. | |
236 if (1 < encoding_types->size()) { | |
237 // Multiple filters were intended to only be used for SDCH (thus far!) | |
238 LogSdchProblem(filter_context, SDCH_MULTIENCODING_FOR_NON_SDCH_REQUEST); | |
239 } | |
240 if ((1 == encoding_types->size()) && | |
241 (FILTER_TYPE_SDCH == encoding_types->front())) { | |
242 LogSdchProblem(filter_context, | |
243 SDCH_SDCH_CONTENT_ENCODE_FOR_NON_SDCH_REQUEST); | |
244 } | |
245 return; | |
246 } | |
247 | |
248 // The request was tagged as an SDCH request, which means the server supplied | |
249 // a dictionary, and we advertised it in the request. Some proxies will do | |
250 // very strange things to the request, or the response, so we have to handle | |
251 // them gracefully. | |
252 | |
253 // If content encoding included SDCH, then everything is "relatively" fine. | |
254 if (!encoding_types->empty() && | |
255 (FILTER_TYPE_SDCH == encoding_types->front())) { | |
256 // Some proxies (found currently in Argentina) strip the Content-Encoding | |
257 // text from "sdch,gzip" to a mere "sdch" without modifying the compressed | |
258 // payload. To handle this gracefully, we simulate the "probably" deleted | |
259 // ",gzip" by appending a tentative gzip decode, which will default to a | |
260 // no-op pass through filter if it doesn't get gzip headers where expected. | |
261 if (1 == encoding_types->size()) { | |
262 encoding_types->push_back(FILTER_TYPE_GZIP_HELPING_SDCH); | |
263 LogSdchProblem(filter_context, SDCH_OPTIONAL_GUNZIP_ENCODING_ADDED); | |
264 } | |
265 return; | |
266 } | |
267 | |
268 // There are now several cases to handle for an SDCH request. Foremost, if | |
269 // the outbound request was stripped so as not to advertise support for | |
270 // encodings, we might get back content with no encoding, or (for example) | |
271 // just gzip. We have to be sure that any changes we make allow for such | |
272 // minimal coding to work. That issue is why we use TENTATIVE filters if we | |
273 // add any, as those filters sniff the content, and act as pass-through | |
274 // filters if headers are not found. | |
275 | |
276 // If the outbound GET is not modified, then the server will generally try to | |
277 // send us SDCH encoded content. As that content returns, there are several | |
278 // corruptions of the header "content-encoding" that proxies may perform (and | |
279 // have been detected in the wild). We already dealt with the a honest | |
280 // content encoding of "sdch,gzip" being corrupted into "sdch" with on change | |
281 // of the actual content. Another common corruption is to either disscard | |
282 // the accurate content encoding, or to replace it with gzip only (again, with | |
283 // no change in actual content). The last observed corruption it to actually | |
284 // change the content, such as by re-gzipping it, and that may happen along | |
285 // with corruption of the stated content encoding (wow!). | |
286 | |
287 // The one unresolved failure mode comes when we advertise a dictionary, and | |
288 // the server tries to *send* a gzipped file (not gzip encode content), and | |
289 // then we could do a gzip decode :-(. Since SDCH is only (currently) | |
290 // supported server side on paths that only send HTML content, this mode has | |
291 // never surfaced in the wild (and is unlikely to). | |
292 // We will gather a lot of stats as we perform the fixups | |
293 if (base::StartsWith(mime_type, kTextHtml, | |
294 base::CompareCase::INSENSITIVE_ASCII)) { | |
295 // Suspicious case: Advertised dictionary, but server didn't use sdch, and | |
296 // we're HTML tagged. | |
297 if (encoding_types->empty()) { | |
298 LogSdchProblem(filter_context, SDCH_ADDED_CONTENT_ENCODING); | |
299 } else if (1 == encoding_types->size()) { | |
300 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODING); | |
301 } else { | |
302 LogSdchProblem(filter_context, SDCH_FIXED_CONTENT_ENCODINGS); | |
303 } | |
304 } else { | |
305 // Remarkable case!?! We advertised an SDCH dictionary, content-encoding | |
306 // was not marked for SDCH processing: Why did the server suggest an SDCH | |
307 // dictionary in the first place??. Also, the content isn't | |
308 // tagged as HTML, despite the fact that SDCH encoding is mostly likely for | |
309 // HTML: Did some anti-virus system strip this tag (sometimes they strip | |
310 // accept-encoding headers on the request)?? Does the content encoding not | |
311 // start with "text/html" for some other reason?? We'll report this as a | |
312 // fixup to a binary file, but it probably really is text/html (some how). | |
313 if (encoding_types->empty()) { | |
314 LogSdchProblem(filter_context, SDCH_BINARY_ADDED_CONTENT_ENCODING); | |
315 } else if (1 == encoding_types->size()) { | |
316 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODING); | |
317 } else { | |
318 LogSdchProblem(filter_context, SDCH_BINARY_FIXED_CONTENT_ENCODINGS); | |
319 } | |
320 } | |
321 | |
322 // Leave the existing encoding type to be processed first, and add our | |
323 // tentative decodings to be done afterwards. Vodaphone UK reportedyl will | |
324 // perform a second layer of gzip encoding atop the server's sdch,gzip | |
325 // encoding, and then claim that the content encoding is a mere gzip. As a | |
326 // result we'll need (in that case) to do the gunzip, plus our tentative | |
327 // gunzip and tentative SDCH decoding. | |
328 // This approach nicely handles the empty() list as well, and should work with | |
329 // other (as yet undiscovered) proxies the choose to re-compressed with some | |
330 // other encoding (such as bzip2, etc.). | |
331 encoding_types->insert(encoding_types->begin(), | |
332 FILTER_TYPE_GZIP_HELPING_SDCH); | |
333 encoding_types->insert(encoding_types->begin(), FILTER_TYPE_SDCH_POSSIBLE); | |
334 return; | |
335 } | |
336 | |
337 std::string Filter::OrderedFilterList() const { | |
338 if (next_filter_) { | |
339 return FilterTypeAsString(type_id_) + "," + | |
340 next_filter_->OrderedFilterList(); | |
341 } else { | |
342 return FilterTypeAsString(type_id_); | |
343 } | |
344 } | |
345 | |
346 Filter::Filter(FilterType type_id) | |
347 : stream_buffer_(nullptr), | |
348 stream_buffer_size_(0), | |
349 next_stream_data_(nullptr), | |
350 stream_data_len_(0), | |
351 last_status_(FILTER_NEED_MORE_DATA), | |
352 type_id_(type_id) {} | |
353 | |
354 Filter::FilterStatus Filter::CopyOut(char* dest_buffer, int* dest_len) { | |
355 int out_len; | |
356 int input_len = *dest_len; | |
357 *dest_len = 0; | |
358 | |
359 if (0 == stream_data_len_) | |
360 return Filter::FILTER_NEED_MORE_DATA; | |
361 | |
362 out_len = std::min(input_len, stream_data_len_); | |
363 memcpy(dest_buffer, next_stream_data_, out_len); | |
364 *dest_len += out_len; | |
365 stream_data_len_ -= out_len; | |
366 if (0 == stream_data_len_) { | |
367 next_stream_data_ = nullptr; | |
368 return Filter::FILTER_NEED_MORE_DATA; | |
369 } else { | |
370 next_stream_data_ += out_len; | |
371 return Filter::FILTER_OK; | |
372 } | |
373 } | |
374 | |
375 // static | |
376 std::unique_ptr<Filter> Filter::InitBrotliFilter(FilterType type_id, | |
377 int buffer_size) { | |
378 std::unique_ptr<Filter> brotli_filter(CreateBrotliFilter(type_id)); | |
379 if (!brotli_filter.get()) | |
380 return nullptr; | |
381 | |
382 brotli_filter->InitBuffer(buffer_size); | |
383 return brotli_filter; | |
384 } | |
385 | |
386 // static | |
387 std::unique_ptr<Filter> Filter::InitGZipFilter(FilterType type_id, | |
388 int buffer_size) { | |
389 std::unique_ptr<GZipFilter> gz_filter(new GZipFilter(type_id)); | |
390 gz_filter->InitBuffer(buffer_size); | |
391 return gz_filter->InitDecoding(type_id) ? std::move(gz_filter) : nullptr; | |
392 } | |
393 | |
394 // static | |
395 std::unique_ptr<Filter> Filter::InitSdchFilter( | |
396 FilterType type_id, | |
397 const FilterContext& filter_context, | |
398 int buffer_size) { | |
399 std::unique_ptr<SdchFilter> sdch_filter( | |
400 new SdchFilter(type_id, filter_context)); | |
401 sdch_filter->InitBuffer(buffer_size); | |
402 return sdch_filter->InitDecoding(type_id) ? std::move(sdch_filter) : nullptr; | |
403 } | |
404 | |
405 // static | |
406 std::unique_ptr<Filter> Filter::PrependNewFilter( | |
407 FilterType type_id, | |
408 const FilterContext& filter_context, | |
409 int buffer_size, | |
410 std::unique_ptr<Filter> filter_list) { | |
411 std::unique_ptr<Filter> first_filter; // Soon to be start of chain. | |
412 switch (type_id) { | |
413 case FILTER_TYPE_BROTLI: | |
414 first_filter = InitBrotliFilter(type_id, buffer_size); | |
415 break; | |
416 case FILTER_TYPE_GZIP_HELPING_SDCH: | |
417 case FILTER_TYPE_DEFLATE: | |
418 case FILTER_TYPE_GZIP: | |
419 first_filter = InitGZipFilter(type_id, buffer_size); | |
420 break; | |
421 case FILTER_TYPE_SDCH: | |
422 case FILTER_TYPE_SDCH_POSSIBLE: | |
423 if (filter_context.GetURLRequestContext()->sdch_manager()) { | |
424 first_filter = InitSdchFilter(type_id, filter_context, buffer_size); | |
425 } | |
426 break; | |
427 default: | |
428 break; | |
429 } | |
430 | |
431 if (!first_filter.get()) | |
432 return nullptr; | |
433 | |
434 first_filter->next_filter_ = std::move(filter_list); | |
435 return first_filter; | |
436 } | |
437 | |
438 void Filter::InitBuffer(int buffer_size) { | |
439 DCHECK(!stream_buffer()); | |
440 DCHECK_GT(buffer_size, 0); | |
441 stream_buffer_ = new IOBuffer(buffer_size); | |
442 stream_buffer_size_ = buffer_size; | |
443 } | |
444 | |
445 void Filter::PushDataIntoNextFilter() { | |
446 IOBuffer* next_buffer = next_filter_->stream_buffer(); | |
447 int next_size = next_filter_->stream_buffer_size(); | |
448 last_status_ = ReadFilteredData(next_buffer->data(), &next_size); | |
449 if (FILTER_ERROR != last_status_) | |
450 next_filter_->FlushStreamBuffer(next_size); | |
451 } | |
452 | |
453 } // namespace net | |
OLD | NEW |